Meir's Monthly Learning Article

This month I want to highlight an article by Kyle Orland, “Cognitive surrender” leads AI users to abandon logical thinking, research finds. This article discusses an emerging concept affecting AI usage: cognitive surrender. Cognitive surrender means adopting AI outputs with minimal scrutiny, overriding intuition and deliberation.

Cognitive surrender feels both scary and unavoidable. People – including me – continually seek to optimize our limited time; the more we offload uninteresting work the more time we can spend on what matters. This pattern predates AI by many decades.

However, a key problem is that AI breaks our expectations of computers based on decades of usage. We are conditioned to computers being completely deterministic, the same inputs give predictable outputs. Thus, out of habit, we treat AI like it is also deterministic. We ask questions or give it prompts and – even if we know better – we start to treat the outputs like they are always correct. Cognitive surrender is the term that defines this degenerate state and will inevitably lead to our collective contracting capability for critical thought.

Lists are where it’s at. Here are a few more thoughts I have about cognitive surrender.

1. I think AI is making us dumber. By “us,” I mean the overall AI-using population, not the people you know who definitely use AI to supercharge their abilities. I said this before. Not everyone agrees.
2. Cognitive surrender is different than the old abacus/calculator line. My early cautious AI position was often challenged, with one friend claiming I would never have replaced the abacus with the calculator. Cognitive surrender is different in that it isn’t about using AI vs. not using AI, instead it’s about how we use AI. The most tempting usage involves offloading all thinking, and that is where the danger lies.
3. We treat AI as if it were 100% reliable. It’s not. Google AI overviews are about 90% correct. I know there are a ton of caveats and model differences and such, however the Google data set is massive enough that the data should not be dominated by outliers. I feel comfortable claiming that today’s AI is roughly 90% correct. Surely this will improve over time, but it will never be 100% because hallucinations are a mathematical inevitability of LLMs. LLMs will never be 100% correct.
4. Use AI, just don’t turn off your brain. AI is powerful, and fun, and amazing. We should use AI, but not in a way we lose our critical thinking skills, or keep the next generations from developing critical thinking skills. Like all parts of the human body, we must use it or lose it. As the linked article data shows, people who use AI to explore topics, learn, and challenge themselves are both improving their own outcomes and maintaining their critical thinking capability. People who offload work to AI and blindly trust the outcomes are succumbing to cognitive surrender.

https://arstechnica.com/ai/2026/04/research-finds-ai-users-scarily-willing-to-surrender-their-cognition-to-llms

“…the group that used AI scored 11.7 percent higher on a measure of their own confidence in their answers, even though the LLM provided wrong answers half the time.”